I think it would be a bit faster (and more accurate) if you use the XRL function (xor).

Such as

top:
mov r1, p1
loop:
mov a, p1
xrl a, r1
JZ loop

; Do what you want here, The different bits
; should be in A

; Probably you should check weather it has been a 1-0
; transition or a 0-1 transition first, if it matters

SJMP top

This probably wont work, but i hope you get the idea. BTW, xrl i think reads from an internal latch, which is the reason that you cannot directly xrl p1 with acc.